Walking requires precise coordination between the legs. This interlimb coordination can be altered using a split-belt treadmill to control the walking speed of each leg independently. Practicing split-belt walking results in the storage of a modified locomotor pattern and, consequently, an after-effect when normal walking conditions are restored. After-effects can be beneficial to some individuals with central nervous system damage (e.g. stroke), raising the exciting possibility that abnormal interlimb coordination can be corrected with adaptive strategies. While this result is promising, recent data have shown that this walking adaptation is quite specific: there is limited transfer to other locomotor tasks (e.g. backward walking) and contexts (e.g. over ground walking). This indicates that the adaptation is caused primarily by alterations of neural circuits specific to the practiced context. In order to optimize rehabilitation techniques, we must facilitate the transfer of walking adaptation to everyday activities. Much remains unknown about how walking adaptation generalizes across various forms of locomotion. To address this issue, we will first generate "generalization functions" to quantify the extent to which an adapted pattern generalizes to different contexts (e.g. different walking speeds, forms). This will be done in both healthy subjects and in stroke survivors, which will allow us to identify differences in generalization functions due to central nervous system damage. Subsequently, we will evaluate different methods to manipulate the boundaries of these functions, resulting in a broader range of generalization. We will also determine whether these techniques can be applied in stroke subjects to facilitate the transfer of walking adaptation to "real- world" tasks like over ground walking. When retraining gait following central nervous system damage, the aim is to train all walking patterns, not just those used for a specific situation. Therefore, we must develop strategies to facilitate the generalization of motor learning during treadmill walking to the forms of walking that will be encountered in the real world. The experiments proposed here will allow us to characterize the dynamic range of locomotor adaptation, which will help us develop strategies to optimally train individuals to improve walking patterns and, ultimately, to improve movement control in general. [unreadable] [unreadable] [unreadable]